Modeling of gas hydrate phase equilibria: Extremely randomized trees and LSSVM approaches

Abstract

This paper involves an innovative computer-based methodology called Extremely Randomized Trees (Extra Trees) to estimate the hydrate formation/dissociation conditions of different gases in the presence of pure water or water-based solutions of salt(s) and/or alcohol(s) for the first time. A smart model is likewise established founded on the least square version of support vector machine (LSSVM). For modeling purpose, an extensive databank containing up more than 1840 experimental data on equilibrium conditions of clathrate hydrate formation/dissociation have been gathered from literature (1940 to 2016). The gathered databank covers extensive variety of pressure-temperature conditions at liquid water-hydrate-vapor or ice-hydrate-vapor equilibria. Error analysis revealed that each of the suggested models replicate the targets with R2% of greater than 96. The forecasts of the evolved Extra Trees and LSSVM models are in good arrangement with corresponding experimental records with the average absolute relative deviation percent (%AARD) identical to 0.04 and 0.32, in that order. Nevertheless, fault examination showed the LSSVM model cannot oppose Extra Trees in expressing the hydrate phase equilibria for the studied gases.